Measuring engineering productivity changed its foundation over the past two years. When a copilot writes 40% of a pull request, counting lines of code became noise accounting. Teams that keep reporting commits and story points steer decisions with data that lost its meaning.
Why lines of code died as a metric
Lines of code, commit counts, and closed story points measure typing. A model like GitHub Copilot produces ten times the volume with equal business impact, so every output metric loses value overnight. Ranking developers by volume compounds the error: the engineer who deletes legacy code and writes dense solutions looks bad next to whoever accepts AI suggestions in bulk.
DORA still works
The four DORA metrics survive the shift because they track flow and collective health: deployment frequency, lead time for changes, change failure rate, and MTTR. None of them grades individuals. All of them answer whether the team ships fast without breaking production. A squad jumping from one deploy per week to one per day after adopting copilots shows a real gain. If the failure rate climbs along, the gain was theater.
New signals worth a dashboard
- Review latency on AI-generated pull requests: when reviewers need three days to read code written in ten minutes, the bottleneck moved.
- Rework rate: the share of merged code reverted or rewritten within 30 days.
- Correlation between AI-assisted changes and production incidents, tracked quarter over quarter.
Surveys beat dashboards
An internal developer experience survey run every quarter reveals where the hours go. Three questions get you started: where did you lose the most time this quarter, which tool blocked you, which process decision slowed you down. The DORA research program polls thousands of engineers per year and finds perception flagging problems before the panels expose them.
The gaming trap
Measuring individual AI usage teaches people to game the number. The engineer who fires off fifty empty prompts looks like the adoption champion. Teams optimize whatever you measure, so pick outcome metrics and behavior follows suit. Publish team-level numbers and skip individual leaderboards altogether.
Where to start
Three DORA metrics plus a three-question survey. Review the numbers every month with the team in the room. Write the results on a one-pager and share it with everyone; shared numbers build partnership, hoarded numbers breed distrust. Keep the survey anonymous or honesty dries up. After one quarter you hold a baseline that tells you where to invest and what to cut.
